What does Generative AI mean?

Generative AI refers to AI systems that create new content: text, images, sound, video, or program code. They continue a given input, adding the most likely next element step by step. Unlike systems that classify or evaluate, they produce a new text or image rather than sorting an existing one into a category.

Text models break the input into tokens and calculate a probability for every possible next token. A temperature setting controls how far the output may deviate from the most probable continuation. Image models take a different route: diffusion methods start from a noise image and remove the noise over many steps, guided by the text description.

Generative AI pays off wherever a first draft can be produced faster than by hand and a review happens anyway. Typical tasks are summarising long documents, rewriting text in plain language, form letters, and code scaffolding. For text with a fixed, legally binding wording, such as a notice of appeal rights, the stored template remains authoritative.

Compared with writing by hand, the focus of the work shifts: with a finished draft in front of them, people correct and shorten rather than formulate from scratch. The same source material can also be turned into several versions, for example a short summary and a plain-language rewrite, without rewriting the text each time.

Because the model chooses its continuation by probability, it produces fluent text even when the necessary facts are missing. Wrong figures, names, and references look exactly like correct ones. Every output must therefore be checked against the source material it was generated from.